Skip to content

roster.md

Latest commit

 

History

History
46 lines (33 loc) · 1.91 KB

roster.md

File metadata and controls

46 lines (33 loc) · 1.91 KB
Raw

Roster — keeping model IDs current

The free/cheap LLM landscape is an open bazaar that rotates monthly: model ids get renamed, :free variants come and go, providers deprecate slots. hermes-max is built so this is always a one-line config edit, never a code change.

How it stays honest

lib/inference/roster.py validates every configured model id at two moments:

  • hm up — warn-only, so a stale id never blocks a start;
  • hm health — the full ROSTER section, one line per slot.

For each provider.slot it:

  1. checks KNOWN_DEPRECATED (a dict populated as models retire);
  2. probes the provider's /models where available (cached 1h, never slows a task): OpenAI-compatible → GET {base}/models; local_vllm → reuse the discovery result; anthropic / cerebras have no /models endpoint → reported unconfirmed (relying on KNOWN_DEPRECATED + a first-call 404).

hm health prints, for example:

  ✓ openrouter.synth_free   moonshotai/kimi-k2.6:free   confirmed
  ✗ groq.synth_oss          openai/gpt-oss-120b         missing
      → NOT in provider /models — update id in inference.yaml

A deprecated or missing slot is a warning, not an error — the system starts anyway and the chain simply uses the next present rung.

Updating a slot

Every model slot in config/inference.example.yaml (or your ~/.hermes-max/inference.yaml) carries a # verified: YYYY-MM-DD comment. When hm health flags one:

  1. find the replacement id on the provider's console / /models;
  2. change the id in the yaml (one line) and update the # verified: date;
  3. hm health re-confirms.

No code change. Ever. This is the maintainability promise: provider knowledge lives in one place (the yaml), behind one seam (lib/inference), with an honest validator. See architecture.md §§16–17 and providers.md.